[BREAKING] Refactor attributes

[h2zero]

Refactor attributes to reduce code duplication and improve maintainability.

• Add attribute base classes to provide common code.
• Add const where possible to functions and parameters.
• NimBLECharacteristic::notify no longer takes a bool is_notification parameter, instead indicate() should be called to send indications.
• NimBLECharacteristic::indicate now takes the same parameters as notify.
• NimBLECharacteristicCallbacks and NimBLEDescriptorCallbacks methods now take const NimBLEConnInfo& instead of non-const.
• NimBLECharacteristic::onNotify callback removed as unnecessary, the library does not call notify without app input.
• NimBLERemoteCharacteristic::getRemoteService now returns a const NimBLERemoteService* instead of non-const.
• Add NimBLEUUID constructor that takes a reference to ble_uuid_any_t.
• NimBLERemoteService::getCharacteristics now returns a const std::vector<NimBLERemoteCharacteristic*>& instead of non-const std::vector<NimBLERemoteCharacteristic*>*
• NimBLERemoteService::getValue now returns NimBLEAttValue instead of std::string
• NimBLEService::getCharacteristics now returns a const std::vector<NimBLECharacteristic*>& instead of a copy of std::vector<NimBLECharacteristic *>.
• Remove const requirement for NimBLEConnInfo parameter in callbacks. Const is unnecessary as the data can't be changed by application code.
• Change NimBLERemoteCharacteristic::getRemoteService to return const pointer.

[finger563]

@h2zero I'm not sure why, but after attempting to upgrade to use this code, I'm finding that my GFPS implementation is no longer working. Compilation works just fine (removing the const from the auth and passkey callbacks), but for whatever reason the onConfirmPIN callback is never being called - and I've found that the event is not even being triggered (the underlying esp-nimble-cpp NimBLEServer::handleGapEvent` is not being called when it should be). I've just started debugging it, but I figured I'd let you know and see if you possibly have some idea of what change might be causing this? The only thing I can think of right now (given that it's the only real change that seems to be coming from this PR) is the change to how attributes are notified/indicated, but I'm not sure...

[h2zero]

Thanks for the report, not sure what would have caused that.

[finger563]

Did some further testing - it's not just GFPS that has broken - the HID service i have is also broken - it seems to be something about how the setValue/indicate/notify change occurred means that the notifications that i send (calling notify) are not being sent to the phone, which manifests as HID inputs not being received by the phone and the GFPS pairing information not being received by the phone either.

[finger563]

Confirmed, even just connecting to the example app, and trying to read the Device Infomation service's characteristics no longer works (testing using the BLE Scanner app on Android).

Edit: these don't always break, they simply break after certain characteristics are read...

Investigation continues...

Edit 2: If i try to read one of the characteristic descriptors, then whatever nimble is doing now cause the BLE Scanner app to crash (oddly no errors on the esp32 side)

[h2zero]

Thanks, did a quick check myself just now. Looks like the call to send the notification returns with ENOT_CONNECTED, I guess the connection handle is wrong somehow.

[finger563]

Looking through the code, I'm not sure where the m_handle for the NimBLECharacteristic is getting set, since it's initialized in the constructor to 0.

Grepping in the code, I'm not finding any places in the code where it's set.